Reliability-based fuzzy clustering ensemble

In the clustering ensemble the quality of base-clusterings influences the consensus clustering. Although some researches have been devoted to weighting the base-clustering, fuzzy cluster level weighting has been ignored, more specifically, they did not pay attention to the role of cluster reliability in the fuzzy clustering ensemble. In this paper, we propose a new fuzzy clustering ensemble framework without access to the features of data-objects based on fuzzy cluster-level weighting. The reliability of each fuzzy cluster is computed based on estimation of its unreliability, and is considered as its weight in the ensemble. The unreliability of fuzzy clusters is estimated by applying the similarity between fuzzy clusters in the ensemble based on an entropic criterion. In our framework, the final clustering is produced by two types of consensus functions: (1) a reliability-based weighted fuzzy co-association matrix is constructed from the base-clusterings and then, a single traditional clustering such as hierarchical agglomerative clustering or K-means is applied over the matrix to produce the final clustering. (2) a new graph based fuzzy consensuses function. The graph based consensus function has linear time complexity in the number of data-objects. Experimental results on various standard datasets demonstrated the effectiveness of the proposed approach compared to the state-of-the-art methods in terms of evaluation criteria and clustering robustness. (c) 2020 Elsevier B.V. All rights reserved.

Automated summary

This paper proposes a new fuzzy clustering ensemble framework based on fuzzy cluster-level weighting to improve clustering results by considering the reliability of clusters. Experimental results demonstrate the effectiveness of the proposed approach compared to state-of-the-art methods in terms of evaluation criteria and clustering robustness.